Email Semantic Search

What it does

Search the mailbox by meaning, not by keyword. Ask for "that receipt from the plumber" or "the thread about the lease" and it finds the right messages even when none of those exact words appear.

How it's built

Every email is turned into a vector embedding and stored in a Qdrant vector database, over 100,000 messages indexed. A query is embedded the same way and matched by cosine similarity, so results rank by semantic closeness rather than literal text. It's available both as this web interface and as a full terminal UI over SSH.

Editor's note: as with most of this, it's a work in progress. The vector-database indexing is real, and works surprisingly well. The web and terminal UIs are still being built out — they're really cool, but they lack important features needed to properly read and manage emails.